{
 "version": "https://jsonfeed.org/version/1.1",
 "title": "Facecast blog",
 "home_page_url": "https://facecast.geenius-os.tech/blog/",
 "feed_url": "https://facecast.geenius-os.tech/blog/feed.json",
 "description": "Practical posts on teleprompters, scripts, camera layouts and recording from the browser, from the team building Facecast.",
 "language": "en",
 "items": [
  {
   "id": "https://facecast.geenius-os.tech/blog/stop-staring-at-the-lens/",
   "url": "https://facecast.geenius-os.tech/blog/stop-staring-at-the-lens/",
   "title": "Read without looking like you are reading",
   "summary": "Why a teleprompter over the camera lens beats notes on the side, and how to set one up so your eyes stay near the viewer.",
   "date_published": "2026-10-17T09:00:00Z",
   "authors": [
    {
     "name": "The Facecast team"
    }
   ],
   "tags": [
    "teleprompter",
    "craft"
   ],
   "content_text": "Everyone has watched the video where the speaker's eyes drift to the left, then back, then to the left again. They are reading notes on another screen, and the viewer feels it before they can say why. Eye contact is a small thing that carries a lot of trust.\nA teleprompter fixes this by putting the words close to the lens. In Facecast the text sits over the video, at the top or bottom of the frame, or beside it. It is never part of the recording, so the viewer sees only you, looking roughly at them.\nPut the text near the lens\nThe closer the text is to the camera, the smaller the eye movement. If you can, put the teleprompter at the top of the frame when your camera is on top of your screen. A narrow column of text moves your eyes less than a wide one. Make the text large enough that you do not lean in: the size runs from 20 to 120 px, and most people are happier bigger than they expect.\nLet the pace do the work\nThe reading pace runs from 60 to 260 words per minute and starts at 140. A calm talk is closer to 130 than 180. Try the pace on the full-screen prompt first, out loud, and change it until you stop rushing the ends of sentences. The line you are on is bright and the rest waits, so you do not have to hunt for your place.\nDo not read everything\nA script does not have to be read word for word. Facecast reads the same text as topics, as bullets or in full. If you know your subject, topics keep you honest and let you sound like yourself. Bullets are a nudge. Full text is for the sentences that must be exact, such as a price or a promise.\nRehearse once\nRead the script out loud before you record, in full mode, and mark the sentences that trip you. Then switch to bullets and say it in your own words. Whichever sounded better is the mode to record in. It costs a minute.\nWhat it does not do\nA teleprompter does not make you a good speaker, and Facecast does not write for you. It has no built-in AI, so the words are yours or those of your own AI, if you connect one. What it does is take away the one thing that looks wrong on camera: the glance away.\nIf you would like to try it, request access and read how to write a script first."
  },
  {
   "id": "https://facecast.geenius-os.tech/blog/write-for-the-ear/",
   "url": "https://facecast.geenius-os.tech/blog/write-for-the-ear/",
   "title": "Write for the ear, not the page",
   "summary": "Short sentences, one idea each, and numbers said out loud: how to write a script that sounds like you when it is read from a teleprompter.",
   "date_published": "2026-10-17T09:00:00Z",
   "authors": [
    {
     "name": "The Facecast team"
    }
   ],
   "tags": [
    "scripts",
    "craft"
   ],
   "content_text": "Text written for the page and text written for the ear are different animals. A sentence that reads well can be a tongue-twister when you say it, and a paragraph that looks tidy can leave a listener behind by the second clause. Scripts for a teleprompter are written for the ear.\nOne idea per sentence\nIf a sentence has two ideas, split it. Short sentences give you room to breathe, and they let the viewer catch up. When you read your script out loud and run out of air, you have found a sentence that is too long.\nSay numbers the way you would say them\nWrite "about two hundred" instead of "~200", and "twenty twenty-six" if that is how you say it. Write the thing you will say, not the thing that looks neat. Dates, prices and names are worth saying out loud once before the take.\nUse the three layers\nFacecast reads a script as topics, bullets or full text. The format is simple: a line starting with # is a topic, a line starting with - is a point to talk through, and every other line is full text. A script written with all three layers gives you a choice on the day.\n# The price\n- three plans, no free tier\n- plans count the takes you keep\nPro is nine euros a month and keeps two hundred takes.\nRead as topics you see "The price". Read as bullets you see the two points. Read in full you see the sentence.\nLeave room\nPut a blank line between paragraphs, because a blank line ends a paragraph and gives your eyes a place to rest. Do not fill every second with words. A pause after an important line works better than another sentence.\nCut the first sentence\nMost scripts begin with a sentence that clears the throat, such as "Hi, in this video I want to talk about". Cut it and start with the thing. The viewer already knows they are in a video.\nCount the length\nWords divided by pace is minutes. A script of 700 words at the default 140 words per minute runs about five minutes. If you wanted three, you have too many words, and the answer is to cut, not to speak faster.\nFor the exact format, see Write a script."
  },
  {
   "id": "https://facecast.geenius-os.tech/blog/topics-bullets-or-every-word/",
   "url": "https://facecast.geenius-os.tech/blog/topics-bullets-or-every-word/",
   "title": "Topics, bullets or every word?",
   "summary": "Three ways to read the same Facecast script, when each works best, and how to move between them as you get more comfortable on camera.",
   "date_published": "2026-10-17T09:00:00Z",
   "authors": [
    {
     "name": "The Facecast team"
    }
   ],
   "tags": [
    "teleprompter",
    "scripts"
   ],
   "content_text": "When people hear "teleprompter" they picture a scroll of text read word for word. That is one way to use it, and it is the least natural. Facecast reads one text three ways, and the interesting part is choosing among them.\nTopics\nTopics show only the headings: a handful of words per part of the talk. It is the right mode when you know your subject and want a map. You say it in your own words, so you sound like yourself, and the map stops you wandering. This is the mode to aim for as you get used to the camera.\nBullets\nBullets show each topic with the points under it. It is the right mode for a talk with facts to cover: three reasons, four steps, a list of features. You get the reminder without the sentence, so you phrase it freshly each time, which sounds like speech because it is.\nEvery word\nFull text shows everything, word for word. It is right when the wording matters: a claim, a legal line, a price, an announcement, a lesson that has to be exact. It is also a good safety net for a nervous first take. Nothing is wrong with using it.\nMove between them\nYou do not have to choose one mode for the whole script. Write all three layers, then read the part you know best as topics and the part with numbers in full. Practise the same script in full, then in bullets, then in topics. The movement from one to the next is how you stop sounding like you are reading.\nThe format\n# Why it matters\n- it saves an afternoon\n- it looks the same every time\nEvery release used to cost us an afternoon.\nThe # line is the topic, the - lines are the bullets, and the plain line is the full text.\nWhat the tool does and does not do\nFacecast does not decide for you and it does not write your script. If you connect your own AI through Geenius OS, it can write the three layers from your notes, and you choose the mode when you record.\nSee rehearse and pace for a routine that works."
  },
  {
   "id": "https://facecast.geenius-os.tech/blog/a-background-that-does-not-distract/",
   "url": "https://facecast.geenius-os.tech/blog/a-background-that-does-not-distract/",
   "title": "A background that does not distract",
   "summary": "Blur, remove or replace the room behind you in Facecast, what runs on your machine, and what is loaded from third parties the first time.",
   "date_published": "2026-10-17T09:00:00Z",
   "authors": [
    {
     "name": "The Facecast team"
    }
   ],
   "tags": [
    "background",
    "craft"
   ],
   "content_text": "Your viewer looks at you for a few minutes, and the room is part of the picture. A pile of laundry or a window with changing light takes attention that should go to what you say. You do not need a studio, but you do need a decision about the background.\nThree modes\nBlur keeps the room but softens it. It is the most forgiving mode, since a slightly wrong edge is hidden in the blur. Remove cuts you out and leaves nothing, which you can fill with a colour, a picture or a looping silent video. Replace is the same with a picture of your own. The edge softness, the stability and the hair detail have sliders, so you can trade a sharp edge for a steady one.\nIf you have a green wall\nA chroma key is there for a green or blue screen, or a colour you pick from the live camera. It is more precise than a model, as long as the light on the screen is even.\nEngines\nSeveral engines make the mask: MediaPipe, BodyPix, MODNet, PP-HumanSeg, and your system's own effect where it exists. They differ in speed, in how they handle hair and in how steady the edge is. Facecast runs them in a worker so the page stays smooth. Try two side by side and keep the one that works with your light.\nWhat is loaded from outside\nThis one matters, so we say it plainly. The engines are third-party models and Facecast does not bundle them. The first time you use one, your browser downloads it from jsDelivr, Google model storage or Hugging Face. Those services can see that your browser asked for the file. The camera picture is processed on your machine and is not sent to them. If that is not acceptable to you, use the blur mode of your system where it exists, or a real wall.\nSome practical rules\nLight your face from the front, not from behind.\nSit a metre from the wall if you can, since blur looks better with depth.\nAvoid shirts the colour of your background.\nRecord a ten-second test and look at the edge of your hair.\nSave a setup you like as a Look, so it is one click next time."
  },
  {
   "id": "https://facecast.geenius-os.tech/blog/slides-and-your-face/",
   "url": "https://facecast.geenius-os.tech/blog/slides-and-your-face/",
   "title": "Slides and your face on the same stage",
   "summary": "How to record a talk with slides in Facecast: a PDF or a Geenius OS Office deck on stage, a face bubble, ink for pointing and chapters for each slide.",
   "date_published": "2026-10-17T09:00:00Z",
   "authors": [
    {
     "name": "The Facecast team"
    }
   ],
   "tags": [
    "layouts",
    "content"
   ],
   "content_text": "Most talks have slides, and most recordings of talks make you choose: show the slides, or show the speaker. A viewer wants both, and a small studio can do that without a second tool.\nPut the slides on stage\nIn Facecast the content shares the stage with you. A PDF opens from a file or a link, and PgUp and PgDn turn the pages. A presentation from Geenius OS Office plays with its transitions and its build animations: → is the next step and ← is back. A document shows as a reading column that moves down a screen at a time.\nChoose the layout for each part\nPress 2 for your face in a bubble over the slide, and 3 for side by side when a slide needs room. Press 4 when a slide should be alone, and 1 for a moment when you are speaking to the viewer. The switch morphs, so it feels like one piece. Each switch is kept as a chapter mark.\nPoint at things\nInk tools have keys, so you do not need the mouse toolbar. A draws an arrow, H a highlight, D a pen and K a laser. C clears. On a slide or a PDF page the ink belongs to that page, so it does not follow you to the next one.\nCapture the moments\nPress S when a slide matters. The shot keeps its time into the take, and you can caption and star it afterwards. Starred shots make a summary card you can end the take with.\nChapters for the viewer\nAfter you keep the take, the layout and source switches become chapters. Copy them as lines into a video description so viewers can jump to a part. The lines look like this:\n00:00 Intro\n02:10 The numbers\n06:45 What to do next\nWhat it does not do\nIt does not edit. A take is a take, so rehearse the order, and use the pause button if you need a breath. It also needs a computer to share a screen: phone browsers have no screen sharing. A PDF or a picture works anywhere, though.\nStart with the content on stage guide."
  },
  {
   "id": "https://facecast.geenius-os.tech/blog/let-your-ai-write-the-script/",
   "url": "https://facecast.geenius-os.tech/blog/let-your-ai-write-the-script/",
   "title": "Let your own AI write the script",
   "summary": "How to ask your own AI to write a three-layer teleprompter script in Facecast through Geenius OS, and what it cannot do for you: it cannot record.",
   "date_published": "2026-10-17T09:00:00Z",
   "authors": [
    {
     "name": "The Facecast team"
    }
   ],
   "tags": [
    "ai",
    "scripts"
   ],
   "content_text": "The hardest part of recording yourself is the words. Facecast has no built-in AI, but if you already work with Claude, ChatGPT or Grok, you can ask it to write the script and put it in your Facecast account through Geenius OS.\nWhat your AI can do\nFacecast gives your AI four tools through Geenius OS. With scripts it can create, read, update and list scripts, and append text to one. With recordings it can read your kept recordings, their marks and their chapters. With shots it can caption, star and order your screenshots. With looks it can save studio setups. It cannot record: recording needs your camera, your screen and your hands, in your browser.\nAsk for the three layers\nThe most useful ask names the format. Facecast reads a script as topics, bullets or every word, so ask for all three.\nUse Geenius OS and my Facecast account. Write a five-minute script for a product demo in three parts. Use a # topic per part, two or three - bullets under it, and then the spoken paragraph. Save it and give me the record link.\nYour AI saves the script and answers with a link. Open it, read it as topics, bullets or full text, and change what does not sound like you.\nMake it yours\nAn AI draft is a starting point. Read it out loud and change every sentence that you would not say. The best use of an AI here is to shape your notes into parts, not to speak for you. Keep the pace in mind: words divided by 140 gives minutes.\nAfter the take\nWhen you have kept a recording, your AI can write the chapters as lines for a description, and caption the screenshots you took. It cannot change the recording file.\nOn which plans\nThe door is on every plan. During early access Facecast is invite-only, and payments are in test mode.\nTo connect, follow use with your AI. The full list of actions is in the reference."
  }
 ]
}